Self-Driving Car Engineer Nanodegree¶
Deep Learning¶
Project: Build a Traffic Sign Recognition Classifier¶
In this notebook, a template is provided for you to implement your functionality in stages, which is required to successfully complete this project. If additional code is required that cannot be included in the notebook, be sure that the Python code is successfully imported and included in your submission if necessary.
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there is a writeup to complete. The writeup should be completed in a separate file, which can be either a markdown file or a pdf document. There is a write up template that can be used to guide the writing process. Completing the code template and writeup template will cover all of the rubric points for this project.
The rubric contains "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. The stand out suggestions are optional. If you decide to pursue the "stand out suggestions", you can include the code in this Ipython notebook and also discuss the results in the writeup file.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode.
Step 0: Load The Data¶
# Load pickled data
import pickle
# TODO: Fill this in based on where you saved the training and testing data
training_file = './TrafficSigns/train.p'
testing_file = './TrafficSigns/test.p'
with open(training_file, mode='rb') as f:
train = pickle.load(f)
with open(testing_file, mode='rb') as f:
test = pickle.load(f)
X_train, y_train = train['features'], train['labels']
X_test, y_test = test['features'], test['labels']
# Debug info
print()
print("(debug) Size of Training & Validation data:",len(X_train))
print("(debug) Size of Test data: ",len(X_test))
Step 1: Dataset Summary & Exploration¶
The pickled data is a dictionary with 4 key/value pairs:
'features'is a 4D array containing raw pixel data of the traffic sign images, (num examples, width, height, channels).'labels'is a 1D array containing the label/class id of the traffic sign. The filesignnames.csvcontains id -> name mappings for each id.'sizes'is a list containing tuples, (width, height) representing the original width and height the image.'coords'is a list containing tuples, (x1, y1, x2, y2) representing coordinates of a bounding box around the sign in the image. THESE COORDINATES ASSUME THE ORIGINAL IMAGE. THE PICKLED DATA CONTAINS RESIZED VERSIONS (32 by 32) OF THESE IMAGES
Complete the basic data summary below. Use python, numpy and/or pandas methods to calculate the data summary rather than hard coding the results. For example, the pandas shape method might be useful for calculating some of the summary results.
Provide a Basic Summary of the Data Set Using Python, Numpy and/or Pandas¶
A) Size & shape analysis of given dataset and generation for validation-set¶
The validation-set is splitted from the original trainings-set by 20%.
### Replace each question mark with the appropriate value.
### Use python, pandas or numpy methods rather than hard coding the results
import numpy as np
from sklearn.model_selection import train_test_split
# Get randomized datasets for training and validation
X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train,test_size=0.2,random_state=0)
# Number of training examples
n_train = len(X_train)
# Number of validation examples
n_validation = len(X_validation)
# Number of testing examples.
n_test = len(X_test)
# What's the shape of an traffic sign image?
image_shape = X_train[0].shape
# How many unique classes/labels there are in the dataset.
n_classes = len(np.unique(y_train))
# print
print()
print("Number of training examples =", n_train)
print("number of validation examples =", n_validation)
print("Number of testing examples =", n_test)
print("Image data shape =", image_shape)
print("Number of classes =", n_classes)
Visualize the German Traffic Signs Dataset using the pickled file(s). This is open ended, suggestions include: plotting traffic sign images, plotting the count of each sign, etc.
The Matplotlib examples and gallery pages are a great resource for doing visualizations in Python.
NOTE: It's recommended you start with something simple first. If you wish to do more, come back to it after you've completed the rest of the sections. It can be interesting to look at the distribution of classes in the training, validation and test set. Is the distribution the same? Are there more examples of some classes than others?
B) Analysis of the distribution of the training-set, validation-set and test-set¶
import matplotlib.pyplot as plt
%matplotlib inline
print ()
print ("Data distributions of training, validation and test sets")
print ()
# Show distribrution of training data
data, number = np.unique(y_train,return_counts=True)
plt.rcParams["figure.figsize"] = [10, 3]
axes = plt.gca()
axes.set_xlim([-1,n_classes])
plt.bar(data, number, tick_label=data, width=0.8, align='center')
plt.title('distribution of training data')
plt.show()
# Show distribution of validation data
data, number = np.unique(y_validation,return_counts=True)
plt.rcParams["figure.figsize"] = [10, 3]
axes = plt.gca()
axes.set_xlim([-1,n_classes])
plt.bar(data, number, tick_label=data, width=0.8, align='center')
plt.title('distribution of validation data')
plt.show()
# Show distribution of test data
data, number = np.unique(y_test,return_counts=True)
plt.rcParams["figure.figsize"] = [10, 3]
axes = plt.gca()
axes.set_xlim([-1,n_classes])
plt.bar(data, number, tick_label=data, width=0.8, align='center')
plt.title('distribution of test data')
plt.show()
C) Visualization of training-set images¶
import random
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
# Print 4 random pics of training data
print ()
print ('Display 4 random training pics examples (title represents the class id)')
# Save random numbers & pics for further steps
pic = np.zeros(4,dtype=int)
plt.figure(figsize=(10, 7))
# Plot images
for i in range(4) :
pic[i] = random.randint(0, len(X_train))
image = X_train[pic[i]]
plt.subplot(4,4,i+1)
plt.imshow(image)
plt.title(y_train[pic[i]])
plt.tight_layout()
plt.show()
# Save pic random numbers
pic_save = np.array(pic,dtype=int)
Step 2: Design and Test a Model Architecture¶
Design and implement a deep learning model that learns to recognize traffic signs. Train and test your model on the German Traffic Sign Dataset.
The LeNet-5 implementation shown in the classroom at the end of the CNN lesson is a solid starting point. You'll have to change the number of classes and possibly the preprocessing, but aside from that it's plug and play!
With the LeNet-5 solution from the lecture, you should expect a validation set accuracy of about 0.89. To meet specifications, the validation set accuracy will need to be at least 0.93. It is possible to get an even higher accuracy, but 0.93 is the minimum for a successful project submission.
There are various aspects to consider when thinking about this problem:
- Neural network architecture (is the network over or underfitting?)
- Play around preprocessing techniques (normalization, rgb to grayscale, etc)
- Number of examples per label (some have more than others).
- Generate fake data.
Here is an example of a published baseline model on this problem. It's not required to be familiar with the approach used in the paper but, it's good practice to try to read papers like these.
Pre-process the Data Set (normalization, grayscale, etc.)¶
Minimally, the image data should be normalized so that the data has mean zero and equal variance. For image data, (pixel - 128)/ 128 is a quick way to approximately normalize the data and can be used in this project.
Other pre-processing steps are optional. You can try different techniques to see if it improves performance.
Use the code cell (or multiple code cells, if necessary) to implement the first step of your project.
A) Preprossing functions: normalize_input, processRGBpics¶
- func. normalize_input, normalizes input image data by pixel - 128)/ 128 => [-1,+1] normalization
- func. processRGBpics, converts color images to grayscale ones and performs a image/pixel normalization (see above)
# Normalize function
def normalize_input(image) :
return (image-128.0)/128
# ProcessRGBpics function coverts to grayscale and normalizes the image (-1,1) style by (pixel - 128)/ 128
def processRGBpics(x):
output = np.average(x, axis=3)
output = np.expand_dims(output, axis=3)
output = normalize_input(output)
return output
B) Pre-processing of training-set, validation-set and test-set data¶
# Process training, validation and test data (grayscale + normalization)
x_training = processRGBpics(X_train)
x_validation = processRGBpics(X_validation)
x_test = processRGBpics(X_test)
C) Visualization of pre-processed training-set images¶
# Plot the processed pics
print ()
print ('Display the 4 random training pics examples from before - processed (title represents the class id)')
plt.figure(figsize=(10, 7))
# Plot processed images
for i in range(len(pic_save)) :
image = np.reshape(x_training[pic_save[i]],(32,32))
plt.subplot(4,4,i+1)
plt.imshow(image,cmap = plt.get_cmap('gray'))
plt.title(y_train[pic_save[i]])
plt.tight_layout()
plt.show()
D) Augmentation - Fake Image Generation¶
In order to increase the training-set random images are generated fromm the original training set. The following random actions are performed on the orignal images in order to create a "fake/new" image:
- width & height movements by +/-10%
- rotation +/-15 degrees
- zoom out/in by 90%/110%
The ImageDataGenerator from the keras enviroment is used to create the fake images.
import keras
from keras.preprocessing.image import ImageDataGenerator
# Setup image manipultion generator from keras
datagen = ImageDataGenerator(rotation_range = 30, width_shift_range = 0.11, height_shift_range = 0.11, zoom_range = 0.2)
# Size of fake data
BATCH = 1000 # batch size generator
RUN = 3 # number of runs
EXIT = int(len(y_train)/BATCH) # exit loops criteria
x_fake =np.zeros((BATCH*EXIT*RUN,32,32,3))
y_fake =np.zeros(BATCH*EXIT*RUN,dtype=np.int)
print ("(debug) Fake images generation started")
for u in range (RUN):
i = 0
k = u * EXIT * BATCH
print ('(debug) loop',u,'of',BATCH*EXIT,'runs ...')
datagen = ImageDataGenerator(rotation_range = 30, width_shift_range = 0.10, height_shift_range = 0.10, zoom_range = 0.2)
for x_, y_ in datagen.flow(X_train, y_train, batch_size = BATCH):
start = k + i * BATCH
end = start + BATCH
y_fake[start:end] = np.copy(y_)
x_fake[start:end] = np.copy(x_)
i = i + 1
if (i>EXIT-1) :
break
print ("(debug) Fake images generated !")
D1) Visualization of generated fake images¶
# Plot 4 random fake images
print ()
print ('Display the 4 random fake pics examples (title represents the class id)')
# Save random numbers & pics for further steps
pic_f = np.zeros(4,dtype=int)
plt.figure(figsize=(10, 7))
# Plot images
for i in range(4) :
pic_f[i] = random.randint(0, len(x_fake))
image = x_fake[pic_f[i]]
plt.subplot(4,4,i+1)
plt.imshow(image)
plt.title(y_fake[pic_f[i]])
plt.tight_layout()
plt.show()
# Save pic random numbers
pic_f_save = np.array(pic_f,dtype=int)
D2) Pre-process augmented images¶
# Process fake image data (grayscale + normalization)
x_training_f = processRGBpics(x_fake)
D3) Visualization of pre-processed augmented images¶
# Plot the processed pics
print ()
print ('Display the 4 random fake pics examples from before - processed (title represents the class id)')
plt.figure(figsize=(10, 7))
# Plot processed images
for i in range(len(pic_f_save)) :
image = np.reshape(x_training_f[pic_f_save[i]],(32,32))
plt.subplot(4,4,i+1)
plt.imshow(image,cmap = plt.get_cmap('gray'))
plt.title(y_fake[pic_f_save[i]])
plt.tight_layout()
plt.show()
D5) Add the augmented images to the training-set¶
# Join fake data to training data (training data = orginal data + fake data)
print ()
print ('(debug) Adding fake data with size =',len(y_fake),'to training data with size=',len(y_train))
print ('(debug) New training set size =',(len(y_fake)+len(y_train)) )
print ()
x_training = np.concatenate((x_training, x_training_f))
y_train = np.concatenate((y_train, y_fake))
D6) Analysis of the distribution of the training-set after adding augmented images¶
# Display distribution of new data
print ()
print ("Data distributions of training after joining in fake data")
print ()
data, number = np.unique(y_train,return_counts=True)
plt.rcParams["figure.figsize"] = [10, 3]
axes = plt.gca()
axes.set_xlim([-1,n_classes])
plt.bar(data, number, tick_label=data, width=0.8, align='center')
plt.title('distribution of training data = orginal + augmented')
plt.show()
D7) Model Architecture based on LeNet architecture¶
from sklearn.utils import shuffle
# shuffle training data
x_training, y_train = shuffle(x_training, y_train)
D7.1) Model functions & Parameters¶
### Define your architecture here.
### Feel free to use as many code cells as needed.
import tensorflow as tf
from tensorflow.contrib.layers import flatten
# training parameters global
EPOCHS = 90
BATCH_SIZE = 128
# Random variables
mu = 0
sigma = 0.1
def conv1_activation(x) :
#SOLUTION: Layer 1: Convolutional. Input = 32x32x1. Output = 30x30x15.
conv1_W = tf.Variable(tf.truncated_normal(shape=(3, 3, 1, 15), mean = mu, stddev = sigma), name = 'conv1_W')
conv1_b = tf.Variable(tf.zeros(15), name = 'conv1_b')
conv1 = tf.nn.conv2d(x, conv1_W, strides=[1, 1, 1, 1], padding='VALID') + conv1_b
# SOLUTION: Activation.
conv1 = tf.nn.relu(conv1)
return conv1
def conv1_subsample(x) :
# SOLUTION: Pooling. Input = 30x30x15. Output = 15x15x15
conv1 = tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
return conv1
def conv2_activation(x):
# SOLUTION: Layer 2: Convolutional. Output = 11x11x16.
conv2_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 15, 16), mean = mu, stddev = sigma),name = 'conv2_W')
conv2_b = tf.Variable(tf.zeros(16), name = 'conv2_b')
conv2 = tf.nn.conv2d(x, conv2_W, strides=[1, 1, 1, 1], padding='VALID') + conv2_b
# SOLUTION: Activation.
conv2 = tf.nn.relu(conv2)
return conv2
def conv2_subsample(x):
# SOLUTION: Pooling. Input = 11x11x16. Output = 5x5x16.
conv2 = tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
return conv2
def fc0_conv1_subsample(x):
# SOLUTION: Pooling. Input = 15x15x15. Output = 7x7x15
sub1 = tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
return sub1
def fc0_net(x,y):
fc0 = tf.concat([flatten(y),flatten(x)],1)
return fc0
def fc1_net(x):
# SOLUTION: Layer 3: Fully Connected. Input = 400 + 735 = 1135. Output = 120.
fc1_W = tf.Variable(tf.truncated_normal(shape=(1135, 120), mean = mu, stddev = sigma), name = 'fc1_W')#
fc1_b = tf.Variable(tf.zeros(120), name = 'fc1_b')
fc1 = tf.matmul(x, fc1_W) + fc1_b
# SOLUTION: Activation.
fc1 = tf.nn.relu(fc1)
return fc1
def fc2_net(x):
# SOLUTION: Layer 4: Fully Connected. Input = 120. Output = 84.
fc2_W = tf.Variable(tf.truncated_normal(shape=(120, 84), mean = mu, stddev = sigma), name = 'fc2_W')
fc2_b = tf.Variable(tf.zeros(84), name = 'fc2_b')
fc2 = tf.matmul(x, fc2_W) + fc2_b
# SOLUTION: Activation.
fc2 = tf.nn.relu(fc2)
return fc2
def LeNet(x):
# SOLUTION: Layer 5: Fully Connected. Input = 84. Output = 43.
fc3_W = tf.Variable(tf.truncated_normal(shape=(84, 43), mean = mu, stddev = sigma), name = 'fc3_W')
fc3_b = tf.Variable(tf.zeros(43), name = 'fc3_b')
logits = tf.matmul(x, fc3_W) + fc3_b
return logits
A validation set can be used to assess how well the model is performing. A low accuracy on the training and validation sets imply underfitting. A high accuracy on the training set but low accuracy on the validation set implies overfitting.
D8) Model Intantiation¶
### Train your model here.
### Calculate and report the accuracy on the training and validation set.
### Once a final model architecture is selected,
### the accuracy on the test set should be calculated and reported as well.
### Feel free to use as many code cells as needed.
x = tf.placeholder(tf.float32, (None, 32, 32, 1)) # color map
y = tf.placeholder(tf.int32, (None))
one_hot_y = tf.one_hot(y, 43) # 43 feautres
rate = 0.001
# model network
conv1_act = conv1_activation(x)
conv1_sub = conv1_subsample(conv1_act)
conv2_act = conv2_activation(conv1_sub)
conv2_sub = conv2_subsample(conv2_act)
fc0_conv1_sub = fc0_conv1_subsample(conv1_sub)
fc0 = fc0_net(fc0_conv1_sub,conv2_sub)
fc1 = fc1_net(fc0)
fc2 = fc2_net(fc1)
logits = LeNet(fc2)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(labels=one_hot_y, logits=logits)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate = rate)
training_operation = optimizer.minimize(loss_operation)
# evaluate
correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))
accuracy_operation = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
D9) Mode Saver & Evaluation function¶
saver = tf.train.Saver()
def evaluate(X_data, y_data):
num_examples = len(X_data)
total_accuracy = 0
sess = tf.get_default_session()
for offset in range(0, num_examples, BATCH_SIZE):
batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]
accuracy = sess.run(accuracy_operation, feed_dict={x: batch_x, y: batch_y})
total_accuracy += (accuracy * len(batch_x))
return total_accuracy / num_examples
D10) Train model¶
from datetime import datetime
# run session
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
num_examples = len(x_training) # independent
v_accuracy = np.zeros(EPOCHS)
v_training = np.zeros(EPOCHS)
print("Training...")
print()
start_time = datetime.now() # start timer
for i in range(EPOCHS):
x_training, y_train = shuffle(x_training, y_train)
for offset in range(0, num_examples, BATCH_SIZE):
end = offset + BATCH_SIZE
batch_x, batch_y = x_training[offset:end], y_train[offset:end]
sess.run(training_operation, feed_dict={x: batch_x, y: batch_y})
training_accuracy = evaluate (x_training, y_train)
v_training[i] = training_accuracy
validation_accuracy = evaluate(x_validation, y_validation)
v_accuracy[i] = validation_accuracy
print("EPOCH {} ...".format(i+1))
print("Training Accuracy = {:.3f}".format(training_accuracy)) # this is perf. inefficient
print("Validation Accuracy = {:.3f}".format(validation_accuracy))
print()
v_test = evaluate (x_test, y_test)
print("Test Accuracy = {:.3f}".format(v_test))
saver.save(sess, './aug-custom-lenet')
print("Run time =",datetime.now()-start_time) # performance
print()
print("Model saved")
Train, Validate and Test the Model¶
Include an exploratory visualization of the dataset¶
A) Visualization of trainings and validation accuracy vs. number of epchos¶
# Plot accuracies
plt.figure(figsize=(26, 10))
plt.subplot(221)
plt.title('Training & Validation Accuracy',fontsize=14)
plt.plot(range(1,EPOCHS+1),(v_accuracy), 'ro')
plt.plot(range(1,EPOCHS+1),(v_accuracy), 'r-')
plt.plot(range(1,EPOCHS+1),(v_training), 'b--')
#plt.text(1,3,'Validation Accuracy',color='red',fontsize=8)
plt.axis([1, EPOCHS, 0, 1.0])
plt.xticks(np.arange(1, EPOCHS+1, 1))
plt.grid()
plt.xlabel('Epochs',fontsize=12)
plt.ylabel('Accuracy',fontsize=12)
plt.text(2,0.2,'Training Accuracy',color='blue',fontsize=12)
plt.text(2,0.15,'Validation Accuracy',color='red',fontsize=12)
plt.text(2,0.1,'Test Accuracy',color='black',fontsize=12)
plt.text(20,0.1,v_test,color='black',fontsize=12)
plt.tight_layout()
plt.show()
Step 3: Test a Model on New Images¶
To give yourself more insight into how your model is working, download at least five pictures of German traffic signs from the web and use your model to predict the traffic sign type.
You may find signnames.csv useful as it contains mappings from the class id (integer) to the actual sign name.
Load and Output the Images¶
A) Helper array to map class label type ID to traffic signs names¶
### Traffic Sign Nsme Mapping function
import numpy as np
with open('signnames.csv') as f:
content = f.read().splitlines()
sign_names = []
# Map to array
for i in range (1,len(content),1) :
name = content[i].split(',')
sign_names.append(name[1])
B) Load 16 traffic sign images from the WWW and visualize these¶
The file name of each image codes the traffic sign class label inside the name: sign_[name].[class type ID].jpg
### Load the images and plot them here.
### Feel free to use as many code cells as needed.
import os
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
# pics location
location = "test_examples.new.good/"
test_dir = os.listdir(location)
pic_features = []
pic_labels = []
# plot pics
plt.figure(figsize=(10, 7))
i = 0
for pics in test_dir :
# read image file
fname = location + pics
image = mpimg.imread(fname)
name = pics.split('.')
# build list
pic_features.append(image)
pic_labels.append(name[1])
# print
#print (pics.shape)
#print (pics)
# show
sub = plt.subplot(5,4,i+1)
plt.imshow(image)
plt.title(pics)
i = i + 1
print ()
print ("(debug)",i, "traffic sign images from the WWW")
print ()
plt.tight_layout()
plt.show()
# pre-process pics for prediction
newfeatures = processRGBpics(np.array(pic_features))
newlabels = np.array(pic_labels,dtype=np.int)
Predict the Sign Type for Each Image¶
C) Predict & visualize sign type prediction for each image (including indirect performance analysis)¶
### Run the predictions here and use the model to output the prediction for each image.
### Make sure to pre-process the images with the same pre-processing pipeline used earlier.
### Feel free to use as many code cells as needed.
import tensorflow as tf
print ()
print ('Sample pics prediction (right pic = original, left pic = used inside model')
print ()
# Run prediction on each pic
i = 0
for pics in newfeatures :
# run predcition with trained model
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
image = np.reshape(pics,(1,32,32,1))
prediction = sess.run(tf.argmax(logits,1), feed_dict={x: image})
# print prediction
print (' SIGN:',sign_names[newlabels[i]] ,' PREDICTION:',sign_names[prediction[0]])
# plot images
plt.subplot(2,2,1)
plt.imshow(pic_features[i])
i = i + 1
plt.subplot(2,2,2)
ima= np.reshape(pics,(32,32))
plt.imshow(ima,cmap = plt.get_cmap('gray'))
plt.tight_layout()
plt.show()
Analyze Performance¶
D) Performance analysis about the 16 traffic sign images from the WWW vs. model (prediction vs. real)¶
# run prediction and accuracy on 16 traffic sign images from the WWW
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
# test accuracy on test-set
test_accuracy = evaluate(x_test, y_test)
print("(debug) Test accuracy = {:.3f}".format(test_accuracy))
# accuracy analysis on 16 traffic signs from WWW
test_accuracy = evaluate(newfeatures, newlabels)
print("(debug) New pics accuracy = {:.3f}".format(test_accuracy))
print("(debug) Prediction perf. analysis for new pics".format(test_accuracy))
# performance analysis on 16 traffic signs from WWW
hits = sess.run(correct_prediction, feed_dict={x: newfeatures, y: newlabels})
prediction = sess.run(tf.argmax(logits,1), feed_dict={x: newfeatures})
# visualize accuracy & performance
# plot analysis vs. images
plt.figure(figsize=(26, 10))
plt.subplot(221)
plt.title('Predicted vs. Real Class Id of Sample Pics')
plt.plot(range(len(newlabels)), newlabels, 'bo')
plt.plot(range(len(newlabels)), newlabels, 'b--')
plt.text(1,5,'Real Class Id',color='blue',fontsize=8)
plt.text(1,3,'Predicted Class Id',color='red',fontsize=8)
plt.plot(range(len(newlabels)), prediction, 'r--')
plt.axis([0, len(newlabels)-1, 0, 43])
plt.xticks(np.arange(0, len(newlabels), 1))
plt.grid()
plt.xlabel('Pic Sample')
plt.ylabel('Class Id')
plt.show()
# plot images
i = 0
plt.figure(figsize=(24, 6))
for pics in pic_features :
if (i>=len(newlabels)) :
break
sub = plt.subplot(4,16,i+1)
sub.set_xticks(())
sub.set_yticks(())
plt.imshow(pics)
plt.title(hits[i],fontsize=30)
i = i + 1
plt.tight_layout()
plt.show()
### Calculate the accuracy for these 5 new images.
### For example, if the model predicted 1 out of 5 signs correctly, it's 20% accurate on these new images.
Output Top 5 Softmax Probabilities For Each Image Found on the Web¶
For each of the new images, print out the model's softmax probabilities to show the certainty of the model's predictions (limit the output to the top 5 probabilities for each image). tf.nn.top_k could prove helpful here.
The example below demonstrates how tf.nn.top_k can be used to find the top k predictions for each image.
tf.nn.top_k will return the values and indices (class ids) of the top k predictions. So if k=3, for each sign, it'll return the 3 largest probabilities (out of a possible 43) and the correspoding class ids.
Take this numpy array as an example. The values in the array represent predictions. The array contains softmax probabilities for five candidate images with six possible classes. tk.nn.top_k is used to choose the three classes with the highest probability:
# (5, 6) array
a = np.array([[ 0.24879643, 0.07032244, 0.12641572, 0.34763842, 0.07893497,
0.12789202],
[ 0.28086119, 0.27569815, 0.08594638, 0.0178669 , 0.18063401,
0.15899337],
[ 0.26076848, 0.23664738, 0.08020603, 0.07001922, 0.1134371 ,
0.23892179],
[ 0.11943333, 0.29198961, 0.02605103, 0.26234032, 0.1351348 ,
0.16505091],
[ 0.09561176, 0.34396535, 0.0643941 , 0.16240774, 0.24206137,
0.09155967]])Running it through sess.run(tf.nn.top_k(tf.constant(a), k=3)) produces:
TopKV2(values=array([[ 0.34763842, 0.24879643, 0.12789202],
[ 0.28086119, 0.27569815, 0.18063401],
[ 0.26076848, 0.23892179, 0.23664738],
[ 0.29198961, 0.26234032, 0.16505091],
[ 0.34396535, 0.24206137, 0.16240774]]), indices=array([[3, 0, 5],
[0, 1, 4],
[0, 5, 1],
[1, 3, 5],
[1, 4, 3]], dtype=int32))Looking just at the first row we get [ 0.34763842, 0.24879643, 0.12789202], you can confirm these are the 3 largest probabilities in a. You'll also notice [3, 0, 5] are the corresponding indices.
E) Analsis of top 5 probabilities of 16 traffic signs images from the WWW¶
### Print out the top five softmax probabilities for the predictions on the German traffic sign images found on the web.
### Feel free to use as many code cells as needed.
# run softmax func. on trained model for the 16 traffic signs from the WWW
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
soft_top5 = sess.run(tf.nn.top_k(tf.nn.softmax(logits),k=5), feed_dict={x: newfeatures})
print ('Top 5 probabilities of the sample pics (first = original, second used @ model)')
print ()
# Plot analysis
#sign_ = np.array(['123456789123456789123456789','123456789123456789123456789','123456789123456789123456789','123456789123456789123456789','123456789123456789123456789'])
#prop_ = np.array([0.0, 0.0, 0.0, 0.0, 0.0])
i = 0
for pics in pic_features :
if (i>=len(newlabels)) :
break
print ('SIGN:',sign_names[newlabels[i]],' RECOGNIZED:',hits[i])
for u in range(5) :
print (' top (',u+1,') {:.20f}'.format(soft_top5[0][i][u]), sign_names[soft_top5[1][i][u]])
#prop_[u] = soft_top5[0][i][u]
#sign_[u] = sign_names[soft_top5[1][i][u]]
plt.subplot(2,2,1)
plt.imshow(pics)
plt.subplot(2,2,2)
ima= np.reshape(newfeatures[i],(32,32))
plt.imshow(ima,cmap = plt.get_cmap('gray'))
#plt.subplot(2,2,3)
#y_pos = np.arange(len(sign_))
#plt.title('Top 5 Probabilities')
#plt.barh(y_pos,prop_,align='center',)
#plt.yticks(y_pos, sign_)
#plt.xticks(prop_)
#plt.xlabel(x_pos)
#plt.xticks(())
#plt.yticks(())
#plt.xlabel('Probability')
plt.tight_layout()
plt.show()
i = i + 1
Note: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
Step 4 (Optional): Visualize the Neural Network's State with Test Images¶
This Section is not required to complete but acts as an additional excersise for understaning the output of a neural network's weights. While neural networks can be a great learning device they are often referred to as a black box. We can understand what the weights of a neural network look like better by plotting their feature maps. After successfully training your neural network you can see what it's feature maps look like by plotting the output of the network's weight layers in response to a test stimuli image. From these plotted feature maps, it's possible to see what characteristics of an image the network finds interesting. For a sign, maybe the inner network feature maps react with high activation to the sign's boundary outline or to the contrast in the sign's painted symbol.
Provided for you below is the function code that allows you to get the visualization output of any tensorflow weight layer you want. The inputs to the function should be a stimuli image, one used during training or a new one you provided, and then the tensorflow variable name that represents the layer's state during the training process, for instance if you wanted to see what the LeNet lab's feature maps looked like for it's second convolutional layer you could enter conv2 as the tf_activation variable.
For an example of what feature map outputs look like, check out NVIDIA's results in their paper End-to-End Deep Learning for Self-Driving Cars in the section Visualization of internal CNN State. NVIDIA was able to show that their network's inner weights had high activations to road boundary lines by comparing feature maps from an image with a clear path to one without. Try experimenting with a similar test to show that your trained network's weights are looking for interesting features, whether it's looking at differences in feature maps from images with or without a sign, or even what feature maps look like in a trained network vs a completely untrained one on the same sign image.
Your output should look something like this (above)
## Visualize your network's feature maps here.
### Feel free to use as many code cells as needed.
# image_input: the test image being fed into the network to produce the feature maps
# tf_activation: should be a tf variable name used during your training procedure that represents the calculated state of a specific weight layer
# activation_min/max: can be used to view the activation contrast in more detail, by default matplot sets min and max to the actual min and max values of the output
# plt_num: used to plot out multiple different weight feature map sets on the same block, just extend the plt number for each new feature map entry
#
# WARNING: changed function outputFeatureMap in order to adapt model structure
# + removed: image_input
# + changed code line: <old> activation = tf_activation.eval(session=sess,feed_dict={x : image_input})
# <new> activation = tf_activation
#
# <old> def outputFeatureMap(image_input, tf_activation, activation_min=-1, activation_max=-1 ,plt_num=1):
#
def outputFeatureMap(tf_activation, activation_min=-1, activation_max=-1 ,plt_num=1):
# Here make sure to preprocess your image_input in a way your network expects
# with size, normalization, ect if needed
# image_input =
# Note: x should be the same name as your network's tensorflow data placeholder variable
# If you get an error tf_activation is not defined it may be having trouble accessing the variable from inside a function
# <old> activation = tf_activation.eval(session=sess,feed_dict={x : image_input})
activation = tf_activation
featuremaps = activation.shape[3]
plt.figure(plt_num, figsize=(15,15))
for featuremap in range(featuremaps):
plt.subplot(6,8, featuremap+1) # sets the number of feature maps to show on each row and column
plt.title('FeatureMap ' + str(featuremap)) # displays the feature map number
if activation_min != -1 & activation_max != -1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin =activation_min, vmax=activation_max, cmap="gray")
elif activation_max != -1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmax=activation_max, cmap="gray")
elif activation_min !=-1:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin=activation_min, cmap="gray")
else:
plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", cmap="gray")
A) Select input image for featuremap analysis, display¶
Variable [image_id] specifies one image of 16 traffic sign images from the WWW
# Image-Featuremaps to display
image_id = 15
print ()
print ('Input image for featuremaps plot (left = original, right = used inside model')
# Visualize input image
plt.subplot(2,2,1)
plt.imshow(pic_features[image_id])
plt.subplot(2,2,2)
im = np.reshape(newfeatures[image_id],(32,32))
plt.imshow(im,cmap = plt.get_cmap('gray'))
plt.tight_layout()
plt.show()
# Prepare image for featuremap display
image = newfeatures[image_id]
image = np.reshape(image,(1,32,32,1))
B) Display featuremaps of model¶
Following featuremaps are displayed:
- conv1 with and without pooling
- conv2 with and without pooling
- 2nd pooling of conv1 which is feed fully connected network fc1 (multi-scale)
# Feature map visualization
with tf.Session() as sess:
# restore session
saver.restore(sess, tf.train.latest_checkpoint('.'))
# Visualize featuremap of conv1
print ()
print ("Featuremap of conv1 network 30x30x15:")
conv1_ = sess.run(conv1_act, feed_dict={x: image})
outputFeatureMap(conv1_)
plt.tight_layout()
plt.show()
print ()
print ("Featuremap of conv1 network after pooling 15x15x15:")
conv1_sub_ = sess.run(conv1_sub, feed_dict={x: image})
outputFeatureMap(conv1_sub_)
plt.tight_layout()
plt.show()
# Visualize featuremap of conv2
print ()
print ("Featuremap of conv2 network 11x11x16:")
conv2_ = sess.run(conv2_act, feed_dict={x: image})
outputFeatureMap(conv2_)
plt.tight_layout()
plt.show()
print ()
print ("Featuremap of conv2 network 5x5x16:")
conv2_sub_ = sess.run(conv2_sub, feed_dict={x: image})
outputFeatureMap(conv2_sub_)
plt.tight_layout()
plt.show()
print ()
print ("Featuremap of conv1 network after 2. pooling 7x7x15 (feed to fc0):")
fc0_conv1_sub_ = sess.run(fc0_conv1_sub, feed_dict={x: image})
outputFeatureMap(fc0_conv1_sub_)
plt.tight_layout()
plt.show()
